DOCUMENT RESUME

ED 105 733 FL 006 754

AUTHOR KDhnc, Takeshi

TITLE Phonology of Some Function Words in Engif,...h
PUB DATE 71
NOTE 25p.; Paper presented at the Summer Institute in
Linguistics at International Christian University
(13th, Tokyo, Japan, 1974)

EDES PRICE MF-$0.76 HC-$1.38 PLUS POSTAGE

DESCRIPTORS Consonants; *Descriptive Linguistics; Determiners
(Languages); English; *Function Words; *Linguistic
Theory; *Phonetics; Phonological Units; *Phonology;
Pronunciation; Synchronic Linguistics; Syntax;
Vowels

ABSTRACT
This paper discusses phonological processes which
assign ultimate phonetic realizations to function words. Stress
patterns of function words are studied along with phonetic variation
between strong and weak forms. The Auxiliary Reduction Rule is
extended to account for the phonetic var.Lation. (Author/AM)
4

Phonology of Some Function Words in English*

Takr'shi Kohn()
Otsuma Women's University

0. In the following chapters phonological processes will

be discussed which assign ultimate phonetic realizations to

function words such as determiners, prepositions, auxil-

iaries and a complementizer. Every function word is marked

as to the lexical feature [M]. A word with [+M], but not

the one with [-M], will undergo subsequent operations of

stress weakening. The stress pattern of each formative is

a given; its stress pattern at a phrase level is (partially)

dependent on its grammatical context rather than its

phonological context. The Auxiliary Reduction Rule (hence-

forth ARR), the only relevant stress weakening rule with

function won:5, must be extended in such a way as to

explain this dependence on grammatical context. Two

possible formulations are in order: the one with gram-

matical categories indicated in the following context and

the one without any. Since they predict different

linguistic facts, the choice of one over the other should

be carefully examined. Tested with a few crucial examples

the labelled formulation of ARR seems to be stronger in its

predictable power than the label-less formulation.

The data which will be examined in tne foilowing are

limited to the Received Pronunciation of British English,

but the data of other dialects of eiLIer British or

American English will be expl_ained in the same way.

t
4.
 2

1. At the outset lot us consider whether it is necessary

to set up stress assignment rules which are applicable to

function words as well as nouns, verbs and adjectives. One

may speculate that function words need not be assigned any

degree of stress at any step of derivation since, unlike

content words, they never occupy stress peaks in an

utterance. Such a speculation, however, will easily turn

out to be false for two reasons. First there exists a

sonority pea% in a multi-syllabic function word. To take

a preposition about as an example, the prominent syllable

falls on the second but not the first. Second, even a

monosyllabic function word is given a stress in some

contexts. It may even take a sentence stress which is

distinct froro emphatic or contrastive stress. Let us

observe the following examples. ( The marks / and,-, stand

for a stressei syllable and an unstressed syllable,

respectively.)

(1) a. l'd like some tea.

b. I'd like some.

(2) a. It's for making the cake rise.

b, What's the powder fOr?

(3) a, You milst leave soon.

b. I suppose you must.

Function words in (lb) , (2b) and (3b) each receive some

degree of stress, with sentence stress on some and must,

whi]e the ones in (la) , (2a) and (3a) lose stress. The

existence of sentences such as in the b-series above will

t,
 3

give a firm ground for setting up stress assignment rules

applicable to function words. The nature of these rules,

however, is not known at present and the clarification of

it is beyond the scope of this paper. Then let us take the

following hypothesis as a basis for examining the stress of

function words.

(4) Hypothesis I: Any function word is given one

stress peak by extended versions of the Main Stress
Rule.

How will the stress pattern of a function word be

modified when it enters a larger cycle? Observe the stress

patterns of the determiners below.

(5) a. a bottle

b. bottle

(6) a. her book

b. JC,an's book

While one and Joan's above have secondary stress at the

Noun Phrase cycle, and he-f have no stress, which can be

ascertained by he presence of the vowel [a) which occurs

only in unstressed positios. The next question will be

what makes the syntactically identical classes phono-

logically distinct as we have just observed above. No

doubt it is due to different derivational processes. In

case of the b-series, after the Main Stress Rule has been

operated on a determiner and a noun, the Nuclear Stress

Rule _is applied to the NP cycle. On the other hand, with

(5a) and (6a), a stress redbction rule seems to have been

responsible for the drastic stress weakening on the

 4

determiners. By the introduction of an extra rule it

became possible to change the secondary-primary stress

contour of the noun phrase into mirus(i. e. weak)-primary.

The applicability of the ARR is a lexical feature of

individual items.

Prepositions ar,d auxiliaries are each divided into

two major classes; those which undergo the operation of

the ARR as in (7a) and (8a) , and those which do not as in

(7b) and (8b).

L.,

(7) a. Let's meet at seven.

b. Let's meet after seven.

(8) a. You must not move.

b. You need not move.

Following is a List of some typical function words which

1
will be dealt with in the following sections, Function

words of Class A in the list are marked with the rule

feature [+M] and those of Class B with [-M]. Any Class A

function word is potentially qualified to undergo the

operation of ARR.

(9) Class A
Determiner a, an, any, (her), (his), Saint,
Sir, some, the, this a' many
Preposition at, but, by, for, from, of, to, into,
onto, toward(s), till, until
Auxiliary am, are, be, been, can, could, do,
does, has, had, have, is, must,
shall, should, was, were, will,
would
Crass B
Determiner thisb, these, that, those, more,
manyb, much, each, all, both,

ti
r
 5

either, neither, two etc., Bill's

etc.
Preposition after, against, in, off, on, over,
round, since, through, under, up,
with, about, before, etc.
Auxiliary dare, did, need, ought

2. Postponing an inquiry into the derivational process

of function words, let us take up their ultimate phonetic

forms at first. Each of the P.M) function words has two

kinds of phonetic forms traditionally termed as "strong

forms" and "weak forms." A strong form appears in a

position which bears some degree of stress, while a weak

form occurs in a position without any recognizable stress

2
at all. Let us compare them below to see if any signif-
3
icant phonetic correspondence lies between them.

(10) Strong Form Weak Form Example

(a) i: I been
u: tr do
Sir
ei Saint
Ea (their)
3: for
will
e (them)
can
from
Is could
A some
(b) h 0 have

As the above table shows, there appear only three vowels

in the list of weak forms in contrast to various vowels in

strong forms. The tense vowel [i:] in the weak form colum

corresponds to its lax counterpart [I], lu:1 to either [U]

or [(9), and the remaining vowels to [a]. Where [a] is

deletable, which in most cases it is, another weak form is

obtained.

The presence or absence of [h] serves as another

index to strong forms vs. weak forms. The consonant is

4
often deleted in front of [a]. Contracted forms of auxil-

iaries result from h-Elision (together with c-)-Elision),

Pronouns such as he, his, him, and her also receive its

application.

1s is clear from the preceding observation, there

exist undeniably fixed patterns of phonetic correspon-

dence between strong forms and weak forms. These patterns,

in fact, cover a broader range than this. They are also

observable between the given segments of any two lexical

formatives which have a common base. To take such a

derivationally related pair sequence and sequential as an

example. When we compare the initial vocalic segments in

both words, we may easily extr;:ct the familiar [i:]-[I]

correspondence. The segment [i:] in sequence is both tense

and stressed, while [m] in sequential is at once lax and

unstressed. A more complete table of correspondence

follows:
,
(11) i: sequence squential
u: cOinpute complt6tion
a: converse conversation
ei expl6in explanation
Ea parent parental
0: inform informtion
Cyrillic Cyril
instrumental instrument
a(1 satire satirical
,
converse converse
A iilUstrious illustrate
h vehicular 0 vehicle
 7

Since the a-Elision is not freely applicable, it may be no

wonder that there happens to be only one such example,

Cyril, on the list. Further instances undergoing its

operation are:

(12) repetition, weaken, rental, faithful, humorous

Both the left-side items and the right-side items on the

table (11) are the same as regards the cyclical stress

assignment rules, but they depart from each other when they

enter word-level phonological rules. While the left-side

items are affected by the Diphthongization Rule which makes

the underlying tense vowels phonetically tense, the Vowel

Shift Rule and the Tensing Rule, the right-side items take
the application of other rules such as Laxing (with ARR as

a primary rule), Vowel Reduction, a-Elision and h-Elision

(as in vehicle). Without any detailed justification, we

assume that all the word-level phonological rules except

ARR are also appli-ablo to function words without necessi-

tating any substantial change in the formulations

originally propo:;ed by Chomsk & Halle (1968). The ARR,

however, needs an extension since the rule in the present

form (cf. (13)) applies to a vowel which may be followed by

other segments within the same (syntactic) word.5

(13) Auxiliary Reduction

I
C (=C
0
) fgstress
L
0( stressl ° V
6 -tense> j

V --,. -stressy
I-tense i irtrstress)
(1 stress] C 0 C [-consli
0
[-stress' V
0
 8

Conditions: 0 = 1, 2, 3, N is weaker than 19,

f is weaker than 2

The rule (13) must be revised so that it can also apply to

a vowel which must be followed by other segments either

within the same word or across the word boundary. To put

it in the concrete, the relevant context to laxen the

underlined vowel in "explanation" is the string of segments

"nation" which follow it in the same word. On the other

hand, in order to laxen the determiner "some," for instance,

as in (la) (but not (1b))by ARR, we shall have to include

as its context segments lying beyond the word boundary;

i. e. the noun "tea." The remainder of this paper is

devoted to an extension of ARR examining determiners,

prepositions, auxiliaries and the complementizer for-to in

that order.

3, 1. Let us observe first the stress of a detei7miner in a

simple noun phrase construction.

(14) I'd like MsOm] e[more] ] Det[tca]NINP

0
(15) I'd like [[[some][more]] 0 ]

(16) I'd like [[[sO'me] 0 l [tea] I

(17) I'd like MsOme] 0 ] 0 1

The formative some is assigned the first stress in (17),

but the minus stress in (14), (15) and (16). The lack of

stress in the latter case can be demonstrated by t1-.e w(,nk

phonetic form fsi7m1 or [sm] . The two kinds of stress above

are supposed to issue from different internal structures of

the noun phrases. While some in (14), (15) and (16) stand
C.
 9

in front of more and tea, the same determiner in (17) does

not. The two accompanying formati\es more and tea here

serve as sister constituents to some in the Noun Phrase.

In order to account for this fact, ARR will have to be

formulated as (18).

(18) V ---). stress / [X

,- , Y]
A
(it) MB
1
-tense
t
+M
l

Conditions: 1. i) B is a sister constituent

of A, and (a) B is dominated by Det. or
(b) B :-= N if A = Det.

As may be apparent from the precedThg rule schema, it

presupposes that the Determiner be a lexical category, and

some and more its immediate constituents. From syntactic

point of view, the Determiner cannot be regarded as a

lexical category since there may appear lower categories

such as Article, Quantifier or the Genitive.. Phonologi-

cally, however, these intermediate categories seem to be

unnecessary. Then there will be no substantial loss if we

omit them by the following Readjustment Rule

(19) The ReaCjustment Rule T:

Truncate any categories below the node Det.

Let us note here in passing that (18) involves

labels A and B. We will refer the ARR such as (18) as "the

Labelled ARR" in contrase with "the Label-less ARR" which

will be presented later for the sake of comparison.

3. 2. In this section we will make a further revision on

(18) so that it may govern prepositions as will. Stress

reduction of a preposition is, as with a determiner,

conditional by the presence of its sister constituent. In

 10

the following examples:

(20) a. He came Ej_fr9:1)1, E[Albertay:,p)pp.

.
b. Whore did he come [(from) p P )
PP.'

(21) a. It's f[forlfmaking the cake rise)).

b. What's the powder [[for] 0 1?

While Prepositions in (20a) and (21a) have their sister

/:oun Phrases after them, ti'ose in (20b) and (21b) do not.

Such syntactic differences invite variations of stress

patterns of Prepositional Phrases. Presumably, then, all

we have ,-:) do to strengthen ARR will be to put an

additional condition on it such as the following:

(22) Condition to (18) : 1. il (c) B = NP if A = P.

3. 3. This section is concerned with the effect of ARR on

auxiliaries. As we shall see, in order to widen the

ferritory for which ARR must be responsible, it will need

a further revision. Under the node Aux in surface

structure are categories keeping the same position as in

the deep structure and grammatical constants which were

moved to that position by transformational rules. The

former include Tense, the Modal, the Perfect, the Progres-

sive and the Passive, and the latter do, the NEG (later

spelled out as not) and the Complementizer. to. Regardless

o'T the heterogeneous sources they behave alike as regards

ARR.
...,

(23) a. They [have Eiwaiting) V for a bus)vp.

---- been) Aux
b, They fhilve 0 ) 0 -

1 1
 11

(24) a. I (have not)[;seen) you for weeks).

h. I not] 0

(25) a. It [has 0 ] [(fallen) on the floor] .

b. It (has 0 ) 0 .

As was indicated by the stress marks on the auxiliaries in

question, all formatives except those in (23b) and (25b),

which are unique predicative elements in respective

sent( aces, result from the operation of ARR. They precede

either another formative in the Auxiliary node or the

associated verb. The formative including the segment which

is affected by ARR and the contextual formative must hold

the rel,itionship "13 is a sister constituent of A," and the

formlitive in question and the associated verb must maintain

such a relationship as "C is a sister constituent of A, and

6
B is directly dominated by C.

The negative particle, which offers an appropriate

cont to a ge,,uine auxiliary in case of the application

of ARR, 1;,a1 either stand by itself (cf. (26a)) or be

aLtche-..! to the preceding formatil'e (often after it was

contracted) (cf. (26b)) .

(26) a. You l) list not] park the car here.

b. (mustn't) pork the car hero.

(26.0 v)sld not ned a further explanation since it is a

parallel exlmple to the previously discussed (24a) . The

newly derived compound formative in (26b) is not subject

to i:P for the sole reason that i, inLluses the nagative

element as its component. From this we can conclude that

 12

for the free operation of ARR the negative particle may

appear separately within the Aux, but must not constitute a

part of the formative in question.

The sequence Aux-Aux (let Aux stand for a formative

dominated by the node Aux) or Aux-V is often interrupted by

a variable such as Adverb (cf. (27) and (28)) or Adverb

plus any other elements (cf. (29)).

(27) I have never drunk hemlock.

(28) They have long been waiting for a bus

(29) have you ever drunk hemlock?

These intervening elements should be omitted from consider-

ation since they function as neutral contexts as to the

applicability of ARR. Their presence or absence do not

alter the stress contour on the auxiliary.

As has been implied in the course of the discussion

so far, suppose conditions on (18) are broadened by

omitting the specification of the relation of the

contextual formative to the auxiliary under consideration,

a wrong result would occur. Instead of the expected

stressed forms as in (30b) and (31b), which are immune from

the operation of ARR, we would inevitably have to admit

reduced forms produced by the unfavoura: le application of

ARR.
v
(30) a. [[I will drive a car] [if I can drive a car
S
1
]s IS
2 0

b. [[T. will 0 ] [if I can 0 ]

 13
... ,..

(31) a. HI will drive a car] S1 [and I must drive a car

1
] S S
2 0
/
b. [(I will 0 ][and I must 0 1].

We would be able to get rid of this unfortunate situation

if we put, among others, a constraint which reads as: The

contextual formative under the node Aux is a sister

constituent to the formative which undergoes ARR.

Before we proceed to make a final revision of ARR

incorporating all the phonological facts we have observed

so far, let us note here that the node Aux has been treated

as a lexical category analogous to the node Det. From the

syntactic representation of the phrase structure such as

(32) :

(32) They [(have] Modal [been] ProgressivelAux

we dropped the specification of the intermediate categories

the Modal and the Progressive. The position we are main-

taining admits two kinds of surface structures: the

syntactic surface structure and the phonological surface

structure. The former is the outcome of all the syntactic

transformations, whereas the latter is produced by the

readjustment rule whose function is to simplify the phrase

structure of the former. Here we propose a readjustment

rule such as the following.

(33) The Readjustment Rule II:

Truncate any categories below the node Aux.

With the help of the preceding rule, it has now become

possible to introduce the final formulation of ARR.

 14

(34) The (Labelled) Auxiliary Reduction

V ---> r-stress // [X YI (([1ql )#)[Z]B
I c
I
-tense I [
+M
Conditions:
1.i) B is a sister constituent of A, and
(a) B is dominated by Det or Aux, or
(b) B = N if A = Det, or
(c) B = NP if A = P. Or
ii) B is directly dominated by a node which
is a sister constituent of A, and
(d) B = V if A = Aux
2. C y B
3. Y does not include n't

3. 4. A short remark is in order on destressing of the

complementizer for-to as in (35).

(35) It is impossible Mairl[JohnlNp[tO1Aux[[

persuadelv Marylvp1s1mp

The pair of constants for-to is one of the markers which

indicate the nounness of a given phrase or clause in which

they appear. If we are permitted to provide the first half

of the complementizers for with the category Preposition,

then subsequent stress reduction process will become

automatic. The application of ARR is constrained in the

same way as genuine prepositions: if a Preposition obeys

a sister Noun Phrase ARR works on it. The constant to, on

the other hand, has been brought into the Aux node as a

substitute for the grammatical formative Tense by the time

ARR applies. Hence, it is given a status equal to other

formatives under the same Aux. The stress on the vowel in

to will be reduced when the item cooccurs with another item

of the Aux or a Verb directly dominated by a category which,

in turn, is a sister to the formative (cf. (.56a)), but it

 15

will not if the necessary conditions are not met (cf. (36b).

(36) a. I'd like [tO] E[take]v a break]vp

Aux
b. I'd like [t6]Aux 0

The identical underlying form is realized as two distinct

phonetic shapes: one as a strong form [tu:] as in (36b),

and the other as a weak form [t] as in (36a). It may be

of some interest that from a phonological point of view the

complementizer for is equated with the preposition for, and

the complementizer to with an auxiliary.

4. So far we have left the question unanswered whether

other formulations of ARR are also possible. We have

chosen the original ARR proposed by Chomsky & Halle in The

Sound Pattern of English as a competing formulation. In

the following discussion we will compare their ARR (or

their modified ARR to be more precise) which does not

jnclude any ca'.egcries in its formulation (i. e. the Label-

less ARR) and ours whLch does include Lheia (I. e. Lhe

Labelled ARR). In anticipating our conclusion, the Label-

less ARR is weaker in descriptive power and includes some

unbearable adhocness of explanation which the Labelled ARR

has not.

Within the framework of Chomsky & Halle (1968), ARR,

as well is other word-level phonological rules, is applied

to a phonological "word," that is, a string limited on both

sides by two boundaries. Prior to the phonologica] rules

readjustment rules are ordered which simplify some parts of

 16

the syntactic surface structure to make a succession of

(phonological) "words." For example, according to Chomsky

& Halle, one of these rules segmentalizes (37) into three

"words," namely, (38a, b, c).7

(37) [s#6Dthelpybook#1m#1Np[vp#was[pp#[pin]p[Np
4[Dan] p[A4un[ Anikely=]A#]A[N#place#]N#Up#]pp

(38) a. ftheitbook#

b. #was4in#an#unnikely#

c. fiplacefl

One thing to note here is that a phonological word does not

always coincide with a syntactic word (in other words,

lexical formative): they may be identical as in (38c) or

the former may be a longer stretch of scunds than the

latter as is (38a) and (38b). Chomsky & Halle (1968)

asserts, among other things, that in (38b), for instance,

the copula was, the preposition in and the determiner an

each form parts of the "word" as proclitics to the major

lexical category Adjective. Although no iemarks were to be

found in their book how to deal with auxiliaries in the

phonological component o.2 English grammar, it will be a

possible natural extension of their contention to treat it

as a proclitic to the accompanying major lexical formative

such as a verb, a noun or an adjective. Hence; to take an

example, (39) will be divided into two words: i. F. (40a)

and (40b).

(39) (#1#(4he#]1/41mp[must have been] Aux" [#fired#]v

"VP"S
 17

(40) a. #he#

b. 4must have been#fired#

In the framework of Chomsky & Halle (1968), no word-level

phonological rules can be applied directly to a sequence

structured as (41) (to repeat a part of (39)).

(41) [must have been] Aux [#[#fired#] V #] VP #] S

Rather, they work on the string (40b), which has had its

internal labelled brackets erased one by one by the last

rule of each cycle. Since syntactic information is no

longer available, it will become impossible to maintain our

conditions 1 i), ii) and 2 of ARR in (34), if we follow

Chomsky & Halle (1968). In order for the original ARR (cf.

(13)) to derive correct stress contour on auxiliaries in

(40b) above, the second case of (13) will be applied first

to the auxiliary been, which includes the last stressed

vowel before the word boundary, then to the immediately

preceding formative have, and lastly to must. As far as

examples such as (40b) are concerned both Chomsky & Halle's

Label-less ARR (13) and our Labelled ARR (34) can bring the

same results. However, the former fails to block the

application in such a case as (42), where there occur no

other predicative elements.

(42) 444you#4must#

The rule (13), then, must be constrained more thoroughly so

that it may not apply to a function word which is the

single formative in a "word." After some revisions it will

take the form as (43).

1
 18

(43) The (Label-less) Auxiliary Reduction

V ----> -stress / ## X Y (#) Z ##
]
[
-tense ] [ 144
Conditions: 1. Z / 0
2. Y does not include n't

The rule (43) by far excel.ls (34) in its simplicity, and we

must frankly admit that there are a considerable number of

cases explicable by it. However, its serious defects will

become prominent when we add the following sentences. (We

will omit irrelevant specifications and let an underline

stand for a "word.")

(44) a. l[HSve)Aux Hyoul)NpHdrunk)v hemlock)))

b. [(Oh,) [Ilve]Aux[[you]]Np]

The correct degree of stress can be brought on the auxil-

iary have by both (43) and (34) but in different ways. By

(43) the vowel in have above is at cnce destressed and

laxened mainly because the auxiliary including it is accom-

panied by another formative, whereas by (34) it undergoes

the same process because the auxiliary has the associated

verb behind it x:Ith permissible optional element between

them. Such a sentence as (44b)lies beyond the descriptive

power of (43). Since the "word" have you in (44b) is

exactly the same as that in (44a), there will be no reason-

able way to block the application in the fn-mer case and to

admit the application in the latter. In contrast, the rule

(34) succeeds in stopping its application to (44b) because

the necessary condition on the associated verb has not been

satisfied.
 19

The inadequacy of the Label-less ARR (43) will be

confirmed by another set of examples.

(45) a. (Has it rained since last month?)

[(Yes,) Hit]j[hAs]
Aux [ [several times] Adv
]vp]
b. (Are they still waiting for a bus ?)
[(Yes,) [[they]][are] Hpatiently] 1
Aux Adv-VP-

c. (Must I leave now?)

,
[(Yes,) Hyoul][must] Aux Has soon as possible
lAdv1VP]

Let us limit our observation to auxiliaries or to those

"words" including them. The rale (43) is incorrectly oper-

ated on these auxiliaries because none appeared alone in

each "word,' while our rule (34) is not because none has a

verb holding a special relation with the formative.

Now that the deficiencies of the Label-less ARR (43)

have been clarified, we will ave a stronger motivation to

replace it by the Labelled ARR (34). In connection with

the validity of the two proposed rules, the adhoc nature

of the "word" in Chomsky & Halle (1968) seems to deserve

some consideration. We may concede that a phonological

"word" need not necessarily coincide with a syntactic

constituent of one kind or another. However, we cannot but

doubt its reality when we find that, in fact, in many cases

it has no verifiable phonetic correlates associated with it.

For, if a "word" is a real phonological unit ,.t must

possibly form a tone group of some size, and the boundary

of any two "words" must potentially be characterized by a

pause. it is totally impossible, however, to insert a

'..)1
 20

pause between the "words" was in an unlikely and place in

(37). Conversely, it would be very hard to do without a

pause (or an intonation break) between the auxiliary and

the adverbial element in each of the sentences (45a, b and

c) even if theoretically they may be supposed to belong to

the same "word." All of these facts together make the

phonological "word" look artificial.

Finally, we may rightly conclude that the Labelled

ARR (34) is the only correct formulation of the rule

accessible at present. The last problem which has been

left untouched is how to keep the labelled brackets unde-

leted till the derivation of a particular string of

segments proceeds to the application of ARR. In Chomsky &

Halle (1968) the innermost paired brackets are erased by

the last rule in each cycle. So, at the time when ARR

begins to operate, such categories are no longer available.

We need not take their line of explanation as the only

possible way to specify the new domain (or cycle) of the

rule. We can achieve the same effect by putting some marks

on the labelled brackets when the operations of all the

rules in a cycle are over instead of erasing the brackets.

This solution will permit us to make use of syntactic

structures at any step of phonological derivation keeping

the principle of the cyclical application of some phonol-

ogical rules intact.

5. The following is a summary of the observations which

2
 21

were made in the preceding sections:

i) Function words such as determiners, prepositions,

auxiliaries and the complementizer for-to must receive the

operation of some stress assignment rule.

ii) Function words are classified into two groups:

the Class A function words undergo a series of stress

weakening rules, whereas those of Class B do not.

iii) Each of the Class A function words has one

strong form and one or more weak forms. These two kinds of

phonetic forms result ultimately from the application or

nonapplication of ARR.

iv) In order to make ARR applicable to function

words, the second case of the Label-less ARR (13) origi-

nally proposed in Chomsky & Halle (1968) must be revised in

such a form as (34) so that it can be operated on a string

which may or may not include a word boundary in its context.

v) Unlike other word-level phonological rules, ARR

is sensitive to syntactic structures of the segments on

which it works as well as phonological environments.

It should be noted that the Labelled ARR (34) is

dependent on syntactic structures more than any other

phonological rules we kn7,w. Of course, in Chomsky & Halle

(1968) we have already encountered similar rules making use

of category names such as Noun, Verb, Adjective and Noun

Phrase for the purpose of marking the domain of each rule.

But our rule (34) depends more heavily on syntactic config-

urations. It has to specify not only grammatical

p
 22

cate9olics of (at least two) formatives containing a

segment which may be affected by the rule and the one

including contextual segments but also syntactic relations

between the formatives. Further investigation is needed on

the influence of the Labelled ARR on the whole system of

English phonology and to determine in a broader perspective

what kinds of syntactic information are funally to be

incorporated in the formulation of phonological rules of

the language.

NOTES

* This paper was presented at The 13th ICU Summer.

Institute in Linguistics held at International Christian
University, Tokyo, in 1974. I should like to express my
thanks to Dr. Eichi Kobayashi, the Chairman of the Division
of Language, ICU, for generously permitting me to copy the
present paper. I am also very grateful to Dr. Erich A.
Berendt for his useful comments on my draft.

1. A few notes on the list (9):

i) Although her and his here belong to determiners, we
do not deal with them directly because they are also pro-
nouns and our primary concern is not about preposit"tons.
ii) Examples of this (cf. (a)) and this (cf. (b))
(a) 1.t's meet, this afternoon.
::

(b) Look at this flower.

iii) Examples of many (cf. (c)) and manyb (cf. (d)) :
(c) How many bro6lers have you?
(d) He speaks many languages fluently.

2. The definition of the strong form and the weak

form can be changed according to the total number of
stresses. Consequently, a theory which distinguishes only
three degrees, i. e. the primary, secondary and weak, may
allow a strong form to occur even in a position bearing a
weak stress. As an example observe the following:
Weak forms occur only in unstressed positions; strong
forms are used chiefly when the word is stressed, but
they also occur in unstressed positions.
(Jones (1960) p. 126)
In order to uncover a deoner relationship between the
 23

strong or weak form and stress, we will reserve the term

"weak form" for a phonetic shape of a formative which has
had its vocalic segment reduced to minus-stress, and "a
strong form" for a shape which is stressed in the broadest
sense.

3. We will exclude the parenthesized items from

consideration.

4. Judging from the fact that although we really

have weak forms such as [1-1'av] and [av] we have not *Ply),
the application of the h-Elision seems to be ordered before
the a-Elision.
5. The rule (13) was quoted from Chomsky & Halle
(1968) p. 240.

6. The following diagrams (i) and (ii) respectively

show the hierarchical structural relationships "B is a
sister constituent of A" and "C is a sister constituent of
A, and B is directly dominated by C."
(i) (ii)

7. The following (37) and (38) were cited from

Chomsky & Halle (1968) p. 368.

REFERENCES

Burt, M. K. From Deep to Surface Structure. New York:

Harper & Row, 1971.

Chomsky, N. and M. Halle. The Sound Pattern of English.

New York: Harper & Row, 1968.

Gimson, A. C. An Introduction to the Pronunciation of

English, 2n1 ed.; London: Edward Arnold, 1970.

Halle, M. and S. J. Keyser. English Stress. New York:

Harper & Row, 1971.

Jones, D. An Outline of English Phonetics, 9th ed.;

Cambridge: W. Heffer & Sons Ltd., 1960.

.
English Pronouncing Dictionary, 13th ed.; London
. M. Dent & Sons Ltd., 1967.
 24

Kenyon, J. S. Amt.!rican Pronunciation, 10th ed.; Ann Arbor:

George Wahr Publishing Company, 1969.

King, H. V. "On Blocking the Rules for Contraction in

English," Linguistic Inquiry, Vol. I, No. 1 (1970)
134-6.

Kingdon, R. The Groundwork of English Intonation. London:

Longmans, 1958.

Kurath, H. A Phonology and Prosody of Modern English.

Ann Arbon : The University of Michigan Press, 1964.

O'Connor, J. D. A Course of English Pronunciation. London:

BBC, 1954.